Most computers work behind the scenes - as embedded systems in many different applications, from cars to industrial machines and medical equipment to household appliances. The future development of networked embedded systems for complex applications depends heavily on artificial intelligence and machine learning.
The software must not only fulfill specific functions, but also meet so-called non-functional requirements such as security, data protection, reliability, accuracy, real-time capability, energy and resource efficiency. Autonomous vehicles and drones, for example, must be protected against failures and unauthorized interference in order to protect people, machines and the environment from harm.

Refinement step by step

"The parameters relevant to the non-functional requirements must be determined early on in the software development process and consistently implemented in further system development," explains Professor Jürgen Becker, spokesperson for the Institute of Information Processing Technology (ITIV) at KIT. The engineering scientist is coordinating the new EU project 'Xandar' (X-by-Construction Design framework for Engineering Autonomous & Distributed Real-time Embedded Software Systems). In 'Xandar', researchers are working in a team with industrial developers to develop a complete toolchain for software development for networked embedded systems based on highly parallel hardware-software platforms with multi-core processors, for example for safety-critical applications in road traffic and aviation.

'Xandar' proposes an innovative approach to this - X-by-Construction. This method is a step-by-step refinement process from specification to code that includes not only the correct function, but also non-functional requirements. "We provide programmers and system developers with a standardized tool chain with which they can control automated hardware-software system integration according to all relevant criteria, especially at an early stage of the design process. With X-by-Construction, we are establishing a new integrated method that significantly improves quality, reduces risks and lowers costs," explains Becker, who heads the Embedded Electronic Systems research area at ITIV.

Eight partners from all over Europe

Xandar' was launched at the beginning of 2021 and is scheduled to run for three years. The European Commission is funding the project under the Horizon 2020 framework program with a total of around 5 million euros. The project consortium includes eight universities, research institutions and companies from all over Europe; KIT is the coordinator. At KIT's ITIV, scientists are developing the core of the tool chain, which incorporates non-functional requirements in all steps of the proposed process - from model-based design and task parallelization to the final code on the platform. Other partners in the Xandar project include the University of Peloponnese, AVN Innovative Technology Solutions Limited, Queen's University Belfast, Vector Informatik GmbH, fentISS, BMW and the German Aerospace Center (DLR).

This article first appeared in our sister magazine Industrial Production.